3,392 karma · joined October 14, 2013
Email me or go to my domain -- freeqaz.com
Personal email: me at freeqaz
----
[ my public key: https://keybase.io/lunasec; my proof: https://keybase.io/lunasec/sigs/4BoTHAkR8TsFV_1S1Rg-Ed_nFD14ocW-gSgKRhhyNIE ]
YC Badge: 0xcb17dfcabdb996705e2859e224d4cdd94d9ae344
I found this link for others that are curious to dig in more. https://www.zdnet.com/article/20-years-later-real-time-linux...
It would be awfully convenient if the way you "store" the nuclear waste is to just leave it in the reactor underground and then build another one!
Immediate problem that comes to mind: Groundwater contamination.
Also it being a pain to build (probably?).
Anybody aware of if anybody has explored this before?
This is the ChatGPT answer that I was able to derive:
> You can read from `/dev/urandom` in a PostgreSQL query using `plperlu`, which allows executing unsafe Perl code. > Create a function to read random bytes:
CREATE EXTENSION plperlu;
CREATE OR REPLACE FUNCTION get_random_bytes(num_bytes int)
RETURNS bytea
LANGUAGE plperlu
AS $$
my $num_bytes = $_[0];
open my $urandom, '<', '/dev/urandom' or die "Cannot open /dev/urandom: $!";
read $urandom, my $bytes, $num_bytes;
close $urandom;
return $bytes;
$$;
SELECT get_random_bytes(16);Taking that thought and thinking about adding "silent" cores is interesting to me. What if your CPU core is actually backed by multiple cores instead to get the "fastest" speed possible? For example imagine if you had say 2 CPU cores that appeared as one and each core would guess the opposite branch of the other (branch prediction) so that it was "right" more of the time.
An interesting thought that had never occurred to me. It's horribly inefficient but for constrained cases where peak performance is all that matters, I wonder if this style of thought would help. ("Competitive Code Execution"?)
Definitely the most impressive level of growth happening at any startup today. I'm curious to see where things head in the future!
Your content is amazing and with some more polish I feel like it would really shine! (Some sentences not flowing quite right is a little confusing for me, reading the GAN deep dive)
Also maybe good for PR?
70B is just a littttle rough trying to run without offloading some layers to the CPU.
What bootloader is this using? Or is it just straight EFI booting?
I helped write a guide a few years back that still is what I do using systemd-boot. https://github.com/lunasec-io/lunasec/blob/master/docs/blog/...
How is Wayland support these days? I love i3 but I know Sway promises to be close enough.
Do you have a rough estimate of how much this cost? I'm curious since I just built my own 2x 3090 rig and I wondered about going EPYC for the potential to have more cards (stuck with AM5 for cheapness though).
All in all I spent about $3500 for everything. I'm guessing this is closer to $12-15k? CPU is around $800 on eBay.
In TechCrunch: AnonMailCoin announces ICO, seeks to provide the Web 4.0 of email
There are plenty of counter-examples to this (writing OSS software, farming your own land with a commune, etc), but in general it's easier to get stuff done when you can throw cash at it, at least in late stage capitalism. It'd be hard to build a company like SpaceX without a boatload of cash!
Does anybody know if they still sell shirts? I would enjoy having one for the memories.
Porkbun is my go to registrar now. And I switched my email to Migadu.
The flip side of this is if the GPU can access the main system memory then I could see this being useful for loading big models with much more efficient "offloading" of layers. Even though bandwidth between GPU->LPDDR5 is going to be slow, it's still faster than what traditional PCI-E would allow.
The caveat here is that I imagine these machines are $$$ and enterprise only. If something like this was brought to the consumer market though I think it would be very enticing.
(If anybody from AMD is reading this, I feel like an architecture like this would be awesome to have. I would love to run Llama 3.1 405b at home and today I see zero path towards doing that for any "reasonable" amount of money (<$10k?).)
Edit: It's at the bottom of the article. These are designed to be meshed together via NVLink into one big cluster.
Makes sense. I'm really curious how the system RAM would be used in LLM training scenarios, or if these boxes are going to be used for totally different tasks that I have little context into.